Demo Abstract:iBookshelf:Accurately Search and Locate Books with an Adaptive and Intelligent Bookshelf Xi ChenT,Lei XieT,Xiaofan Jiang,Sanglu Lu',and Daoxu Chen State Key Laboratory for Novel Software Technology,Nanjing University,Nanjing,China ￥Intel Labs China hawkxc@dislab.nju.edu.cn,Ixie@nju.edu.cn,fred.jiang@intel.com,{sanglu,cdx@nju.edu.cn Abstract It is a tedious task to search and locate a specific book from massive number of books arbitrarily placed in a book- shelf.In this paper,we demonstrate iBookshelf,a system which allows users to quickly search and accurately locate books in the bookshelf,by leveraging a passive RFID sys- tem.By deploying a number of reference tags on the book- shelf,we are able to perform localization based on the simi- larities in received signal strength,and effectively offset the impact from the ambient noises and interferences.We de- ploy and evaluate our system in a real 3mx2.5m bookshelf. Figure 1.Real image of iBookshelf and show that users are able to locate the book from our Android-based application with 85%accuracy. Keywords resilient to environmental changes by adaptively conduct- Localization.Search.Bookshelf.Passive RFID ing self-calibration,and locates the book with 85%accu- 1 racy in a real-time approach.In this system,each book is Introduction attached with a passive RFID tag,which costs a few cents In our daily lives,most people will find that it is a te- per tag.The passive RFID reader obtains the information by dious task to search and locate a specified book from a large resolving the backscattered signals from passive tags,there- volume of books,which are often arbitrarily placed in the fore,the received signal strength is very sensitive to the dis- bookshelf.The traditional indoor localization techniques are tance,which improves localization accuracy.According to mainly based on WiFi[5],sound[3]as well as active RFID keywords stored in the tags,the system indexes the corre- [2],among which the highest accuracy is 1m~2m.Howev- sponding information of the books and supports information er,these techniques are not suitable for locating a book in the retrieval similar to using the Google search bar.By means bookshelf due to the relative low accuracy in localization of the reference tags on the bookshelf,the system is able to Meanwhile,current research on "smart"bookshelves[1][4] accurately locate the book based on the similarities in the re- either seek to locate books through optical localization using ceived signal strength,and effectively offset the impact from integrated LEDs or require calibration through a laborious the ambient noises and interferences.In our demonstration, training phase. the books are placed into a bookshelf with 4x4 grids,as In this paper,we demonstrate iBookshelf,a cost-effective shown in Fig.1.Users can use the custom-designed Android- system which allows users to search and accurately locate based tablet embedded in the bookshelf to search and locate the books in the bookshelf,by leveraging the passive RFID the specified books. system.Different from the previous work,by deploying a number of reference tags on the bookshelf,our system is 2 System Design Our system is composed of three main parts:(1)data ac- Video demo is available at http://cs.nju.edu.cn/lxie/ibookshelf.htm quisition module,(2)indexing/search module,and (3)local- ization module.Fig.2 shows the design model of iBookshelf. The bookshelf is composed of 16 grids with 4 columns and 4 rows,the size of each grid is 60cm x 75cm.As Fig.2 il- lustrates,the data acquisition module includes the reader as well as the antennas.In current implementation,we use the Alien-9900 reader and Alien-9611 linear antenna with a di- rectional gain of 6dB.We deploy the reader on the bottom of Copyright is held by the author/owner(s).SenSys'12,November 6-9,2012,Toronto, Canada.AM978-1-4503-1169-411/12 the bookshelf,while placing 2 antennas on both sides of the bookshelf.The indexing/search module and the localization ACM module are implemented in the tablet PC,which communi-
Demo Abstract: iBookshelf: Accurately Search and Locate Books with an Adaptive and Intelligent Bookshelf Xi Chen† , Lei Xie† , Xiaofan Jiang‡ , Sanglu Lu† , and Daoxu Chen† †State Key Laboratory for Novel Software Technology, Nanjing University, Nanjing, China ‡ Intel Labs China hawkxc@dislab.nju.edu.cn, lxie@nju.edu.cn, fred.jiang@intel.com, {sanglu,cdx}@nju.edu.cn Abstract It is a tedious task to search and locate a specific book from massive number of books arbitrarily placed in a bookshelf. In this paper, we demonstrate iBookshelf1 , a system which allows users to quickly search and accurately locate books in the bookshelf, by leveraging a passive RFID system. By deploying a number of reference tags on the bookshelf, we are able to perform localization based on the similarities in received signal strength, and effectively offset the impact from the ambient noises and interferences. We deploy and evaluate our system in a real 3m×2.5m bookshelf, and show that users are able to locate the book from our Android-based application with 85% accuracy. Keywords Localization, Search, Bookshelf, Passive RFID 1 Introduction In our daily lives, most people will find that it is a tedious task to search and locate a specified book from a large volume of books, which are often arbitrarily placed in the bookshelf. The traditional indoor localization techniques are mainly based on WiFi[5], sound[3] as well as active RFID [2], among which the highest accuracy is 1m∼2m. However, these techniques are not suitable for locating a book in the bookshelf due to the relative low accuracy in localization. Meanwhile, current research on “smart” bookshelves[1][4] either seek to locate books through optical localization using integrated LEDs or require calibration through a laborious training phase. In this paper, we demonstrate iBookshelf, a cost-effective system which allows users to search and accurately locate the books in the bookshelf, by leveraging the passive RFID system. Different from the previous work, by deploying a number of reference tags on the bookshelf, our system is 1Video demo is available at http://cs.nju.edu.cn/lxie/ibookshelf.htm Copyright is held by the author/owner(s). SenSys ’12, November 6-9, 2012, Toronto, Canada. ACM 978-1-4503-1169-4/11/12 ACM Figure 1. Real image of iBookshelf resilient to environmental changes by adaptively conducting self-calibration, and locates the book with 85% accuracy in a real-time approach. In this system, each book is attached with a passive RFID tag, which costs a few cents per tag. The passive RFID reader obtains the information by resolving the backscattered signals from passive tags, therefore, the received signal strength is very sensitive to the distance, which improves localization accuracy. According to keywords stored in the tags, the system indexes the corresponding information of the books and supports information retrieval similar to using the Google search bar. By means of the reference tags on the bookshelf, the system is able to accurately locate the book based on the similarities in the received signal strength, and effectively offset the impact from the ambient noises and interferences. In our demonstration, the books are placed into a bookshelf with 4×4 grids, as shown in Fig.1. Users can use the custom-designed Androidbased tablet embedded in the bookshelf to search and locate the specified books. 2 System Design Our system is composed of three main parts: (1) data acquisition module, (2) indexing/search module, and (3) localization module. Fig.2 shows the design model of iBookshelf. The bookshelf is composed of 16 grids with 4 columns and 4 rows, the size of each grid is 60cm × 75cm. As Fig.2 illustrates, the data acquisition module includes the reader as well as the antennas. In current implementation, we use the Alien-9900 reader and Alien-9611 linear antenna with a directional gain of 6dB. We deploy the reader on the bottom of the bookshelf, while placing 2 antennas on both sides of the bookshelf. The indexing/search module and the localization module are implemented in the tablet PC, which communi-
cate with the data acquisition module through WiFi. 3 Demonstration We start by randomly place books into the bookshelf.The LD tag user is invited to search a book by entering the keywords in- to the search box in the tablet embedded in the bookshelf. By clicking on one of the results,information such as the abstract and cover of the book are crawled from the Internet and illustrated on the user interface.The user can further lo- cate the book's position by clicking the"location"button,i- Bookshelfreturns the location graphically as shown in Fig.4, along with less probable locations and the respective proba- 4a- X-axis bilities.Due to space limitation,we can show a video of the Figure 2.The design model of iBookshelf system at the demo,and present a smaller live version 2.1 Indexing and Searching The indexing/search module is working as follows:once a book with tag attached is placed into the bookshelf,the tag'ID and keyword information is sensed by the data ac- quisition module.Then,the keywords are incorporated into the search index,meanwhile,the web crawlers use these key- words to crawl more detail information from web sites such as Wikipedia in the Internet.The crawled data are further Theya stored and indexed.When a user wants to search a book. The x axis he/she enters the keyword and the system returns the search (a)The similarity between the target(b)The probability of each grid for tag and the reference tags the target book to be located to results by fast retrieving data from the index 2.2 Localization Figure 3.Experiment results:the target book is deployed in Grid (3,2)in the X-Y axis among the 4 x 4 grids The objective of the localization module is to accurately locate the target tags in a real-time approach.In order to offset the impact from the ambient noises and interferences. we uniformly deploy 15 reference tags inside the grids,as shown in Fig.2.The position of each reference tag in X-Y axis is fixed and known in advance.We propose an algorithm to estimate the target tag's position,as shown in Alg.1. Algorithm 1 Localization Algorithm 1:Set the reader to issue m query cycles from each of the 4 antennas. 2:Get the signal strength vector of the target tag as V={s1,s2,53,s4),sj is the average signal strength of the target tag perceived on antennas / Figure 4.The user interface of iBookshelf where j∈[l,4. 3:Get the signal strength vector of the reference tags as= 4 Acknowledgments {s.1,2,s.3,s.4},where i∈[l,15. This work is supported by the National Basic Research 4:for i 1,15]do Program of China(973)under Grant No.2009CB320705: Compute the similarity between and Vi as follows the National Natural Science Foundation of China under V.V mK,闭=个 =1 Grant No..61100196,61073028,61021062:the JiangSu V=了V= Natural Science Foundation under Grant No.BK2011559. References 6 end for 7:Sort the value of sim(.)in decreasing order.Find the first k refer- [1]S.Inoue,Y.Nohara,T.Masaki,and K.Sakuragawa.Location recogni- tion in rfid bookshelves.IEICE Transactions,pages 1147-1152,2011. ence tags with the position (x)according to sim(V.V). 8:Compute the position of the target tag [2]L.M.Ni,Y.Liu,Y.C.Lau,A.P.Patil,and A.P.Patil.Landmarc: Indoor location sensing using active rfid.Wireless Nenorks,pages (.y)()-w.here.= 1/(1-sim(V,)+e) 701-710,2004 1 I-m(化月+g>0叭， [3]C.Peng,G.Shen,Y.Zhang,Y.Li,and K.Tan.Beepbeep:a high accuracy acoustic ranging system using cots mobile devices.In Proc. According to the localization algorithm,suppose we place ofACM SenSys,2007. a target book in Grid (3,2)in the X-Y axis,Fig.3(a)shows the [4]A.Sample,C.Macomber,L.-T.Jiang,and J.R.Smith.Optical local- similarities between this target tag and the reference tags in ization of passive uhf rfid tags with integrated leds.In Proc.of /EEE terms of the received signal strength,Fig.3(b)further shows RFID,2012. the probability of each grid that the target book is located to, [5]M.Youssef,M.Mah,and A.Agrawala.Challenges:device-free passive on average the target book is located to the right grid with localization for wireless environments.In Proc.of ACM MobiCom, 85%accuracy 2007
cate with the data acquisition module through WiFi. RFID Reader (Alien-9900) Antenna Antenna Books attached with RFID tag Antenna Antenna Tablet PC Reference RFID tags X-axis Y-axis (1,3) (2,3) (3,3) (4,3) (5,3) (1,2) (2,2) (3,2) (4,2) (5,2) (1,1) (2,1) (3,1) (4,1) (5,1) 1 2 3 4 1 1 2 3 4 Figure 2. The design model of iBookshelf 2.1 Indexing and Searching The indexing/search module is working as follows: once a book with tag attached is placed into the bookshelf, the tag’ ID and keyword information is sensed by the data acquisition module. Then, the keywords are incorporated into the search index, meanwhile, the web crawlers use these keywords to crawl more detail information from web sites such as Wikipedia in the Internet. The crawled data are further stored and indexed. When a user wants to search a book, he/she enters the keyword and the system returns the search results by fast retrieving data from the index. 2.2 Localization The objective of the localization module is to accurately locate the target tags in a real-time approach. In order to offset the impact from the ambient noises and interferences, we uniformly deploy 15 reference tags inside the grids, as shown in Fig.2. The position of each reference tag in X-Y axis is fixed and known in advance. We propose an algorithm to estimate the target tag’s position, as shown in Alg. 1. Algorithm 1 Localization Algorithm 1: Set the reader to issue m query cycles from each of the 4 antennas. 2: Get the signal strength vector of the target tag as V = {s1,s2,s3,s4}, sj is the average signal strength of the target tag perceived on antennas j, where j ∈ [1,4]. 3: Get the signal strength vector of the reference tags as Vi = {si,1,si,2,si,3,si,4}, where i ∈ [1,15]. 4: for i ∈ [1,15] do 5: Compute the similarity between V and Vi as follows: sim(V,Vi) = V ·Vi |V|·|Vi | = ∑ 4 j=1 si, j ·sj q ∑ 4 j=1 s 2 j · q ∑ 4 j=1 s 2 i, j . 6: end for 7: Sort the value of sim(V,Vi) in decreasing order. Find the first k reference tags with the position (xi ,yi) according to sim(V,Vi). 8: Compute the position of the target tag (x,y) = k ∑ i=1 (xi ,yi)·wi ,here,wi = 1/(1−sim(V,Vi) +ε) ∑ k i=1 1/(1−sim(V,Vi) +ε) (ε > 0). According to the localization algorithm, suppose we place a target book in Grid (3,2) in the X-Y axis, Fig.3(a) shows the similarities between this target tag and the reference tags in terms of the received signal strength, Fig.3(b) further shows the probability of each grid that the target book is located to, on average the target book is located to the right grid with 85% accuracy. 3 Demonstration We start by randomly place books into the bookshelf. The user is invited to search a book by entering the keywords into the search box in the tablet embedded in the bookshelf. By clicking on one of the results, information such as the abstract and cover of the book are crawled from the Internet and illustrated on the user interface. The user can further locate the book’s position by clicking the “location” button, iBookshelf returns the location graphically as shown in Fig.4, along with less probable locations and the respective probabilities. Due to space limitation, we can show a video of the system at the demo, and present a smaller live version. 1 2 3 4 5 1 2 3 0 0.5 1 The value of y The value of x The value of similarity (a) The similarity between the target tag and the reference tags 1 2 3 4 1 2 3 4 0 20 40 60 80 100 The x axis The y axis The detection ratio (%) (b) The probability of each grid for the target book to be located to Figure 3. Experiment results: the target book is deployed in Grid (3,2) in the X-Y axis among the 4 × 4 grids Figure 4. The user interface of iBookshelf 4 Acknowledgments This work is supported by the National Basic Research Program of China (973) under Grant No. 2009CB320705; the National Natural Science Foundation of China under Grant No. 61100196, 61073028, 61021062; the JiangSu Natural Science Foundation under Grant No. BK2011559. References [1] S. Inoue, Y. Nohara, T. Masaki, and K. Sakuragawa. Location recognition in rfid bookshelves. IEICE Transactions, pages 1147–1152, 2011. [2] L. M. Ni, Y. Liu, Y. C. Lau, A. P. Patil, and A. P. Patil. Landmarc: Indoor location sensing using active rfid. Wireless Networks, pages 701–710, 2004. [3] C. Peng, G. Shen, Y. Zhang, Y. Li, and K. Tan. Beepbeep: a high accuracy acoustic ranging system using cots mobile devices. In Proc. of ACM SenSys, 2007. [4] A. Sample, C. Macomber, L.-T. Jiang, and J. R. Smith. Optical localization of passive uhf rfid tags with integrated leds. In Proc. of IEEE RFID, 2012. [5] M. Youssef, M. Mah, and A. Agrawala. Challenges: device-free passive localization for wireless environments. In Proc. of ACM MobiCom, 2007